the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
High-resolution predictions of ground ice content for the Northern Hemisphere permafrost region
Abstract. Ground ice content of the Arctic soils largely dictates the effects of climate change-induced permafrost degradation and top ground destabilization. The current circumarctic information on ground ice content is overly coarse for many key applications, including assessments of hazards to Arctic infrastructure, while detailed data are restricted to very few regions. This study aims to address these gaps by presenting spatially comprehensive data on pore and segregated ground ice content across the Northern Hemisphere permafrost region at a 1-km resolution. First, ground ice content datasets (n=437 and 380 1-km grid cells for volumetric and gravimetric ice content, respectively) were compiled from field observations over the permafrost region. Spatial estimates of ground ice content in the near-surface permafrost north of the 30th parallel north were then produced by relating observed ground ice content to physically relevant environmental data layers of climate, soil, topography, and vegetation properties using a statistical modelling framework. The produced data show that ground ice content varies substantially across the permafrost region. The highest ice contents are found on peat-dominated Arctic lowlands and along major river basins. Low ice contents are associated with mountainous areas and many sporadic and isolated permafrost regions. The modelling yields relatively small prediction errors (a mean absolute error of 13.6 % volumetric ice content) over evaluation data and broadly congruent spatial distributions with earlier regional-scale studies. The presented data allow the consideration of ground ice content in various geomorphological, ecological, and environmental impact assessment applications at a scale that is more relevant than previous products. The produced ground ice data are available in the supplement for this study and at Zenodo https://doi.org/10.5281/zenodo.7009875 (Karjalainen et al., 2022).
- Preprint
(4980 KB) - Metadata XML
-
Supplement
(100 KB) - BibTeX
- EndNote
Status: closed
- RC1: 'Comment on essd-2022-144', Hugh O'Neill, 12 Oct 2022
-
RC2: 'Comment on essd-2022-144', Anonymous Referee #2, 31 Oct 2022
The data set provides very important information about near-surface (upper 5m) ground ice content for the Northern hemisphere. Such data are today available only very course but would be urgently needed for climate change projections and climate change impact studies related to permafrost thaw. The paper is well written and presented.
My main critics is that the model developed to estimate ground ice content is purely statistical based on scattered field measurements, and does little involve the physical processes behind the ground ice content and its evolution over a range in time scales. Statistical models are weak (or invalid) in representing conditions that are not or little represented in the training data. It is unclear to me to what extent the ice content data are biased or clustered in various aspects (topography, climate, groundtype, ice formation history, etc.). The authors mention actually a bias to ice-rich conditions. Such biases are not included in the uncertainty quantification. Similarly, it is unclear to what extent and why the input variables used to drive the statistical models represent the processes associated with ground ice formation and content. Given these uncertainties, the formal uncertainties given will not reflect the full reliability of the new data set, and it becomes thus not clear for which applications it could be used, and for which rather not. I can imagine users might apply your data in a way not justified by their validity and accuracy.
I understand my comments are likely not easy to include in the study. A more careful description of the non-formalized sources of uncertainties would be needed and an attempt to quantify these. Should areas be excluded that extrapolate outside of the conditions covered by the training data the parameter space?
Minor comments:
Though well-expected for expert readers, the title (or at least the abstract) should clearly state the the paper talks about the upper 5 m.
Lines 19 and 336: I wouldn't say the data "show" that... It is a statistical model. The produced data contain ... or similar
Line 72: You refer to SROCC, right? This is "an" IPCC special report, there are several other ones.
Fig 3 a: black dots on dark pink ground difficult to recognize.
Fig C4: Possible to have the input data in the background for each panel? For instance as colourcoded historgramme (as in Fig 5, perhaps in greyscale)?
Â
Citation: https://doi.org/10.5194/essd-2022-144-RC2
Status: closed
- RC1: 'Comment on essd-2022-144', Hugh O'Neill, 12 Oct 2022
-
RC2: 'Comment on essd-2022-144', Anonymous Referee #2, 31 Oct 2022
The data set provides very important information about near-surface (upper 5m) ground ice content for the Northern hemisphere. Such data are today available only very course but would be urgently needed for climate change projections and climate change impact studies related to permafrost thaw. The paper is well written and presented.
My main critics is that the model developed to estimate ground ice content is purely statistical based on scattered field measurements, and does little involve the physical processes behind the ground ice content and its evolution over a range in time scales. Statistical models are weak (or invalid) in representing conditions that are not or little represented in the training data. It is unclear to me to what extent the ice content data are biased or clustered in various aspects (topography, climate, groundtype, ice formation history, etc.). The authors mention actually a bias to ice-rich conditions. Such biases are not included in the uncertainty quantification. Similarly, it is unclear to what extent and why the input variables used to drive the statistical models represent the processes associated with ground ice formation and content. Given these uncertainties, the formal uncertainties given will not reflect the full reliability of the new data set, and it becomes thus not clear for which applications it could be used, and for which rather not. I can imagine users might apply your data in a way not justified by their validity and accuracy.
I understand my comments are likely not easy to include in the study. A more careful description of the non-formalized sources of uncertainties would be needed and an attempt to quantify these. Should areas be excluded that extrapolate outside of the conditions covered by the training data the parameter space?
Minor comments:
Though well-expected for expert readers, the title (or at least the abstract) should clearly state the the paper talks about the upper 5 m.
Lines 19 and 336: I wouldn't say the data "show" that... It is a statistical model. The produced data contain ... or similar
Line 72: You refer to SROCC, right? This is "an" IPCC special report, there are several other ones.
Fig 3 a: black dots on dark pink ground difficult to recognize.
Fig C4: Possible to have the input data in the background for each panel? For instance as colourcoded historgramme (as in Fig 5, perhaps in greyscale)?
Â
Citation: https://doi.org/10.5194/essd-2022-144-RC2
Data sets
Ground ice content predictions for the Northern Hemisphere permafrost region at 1-km resolution, version 1.1 Karjalainen, Olli; Aalto, Juha; Kanevskiy, Mikhail Z.; Luoto, Miska; Hjort, Jan https://doi.org/10.5281/zenodo.7009875
Viewed
HTML | XML | Total | Supplement | BibTeX | EndNote | |
---|---|---|---|---|---|---|
1,373 | 640 | 64 | 2,077 | 134 | 84 | 79 |
- HTML: 1,373
- PDF: 640
- XML: 64
- Total: 2,077
- Supplement: 134
- BibTeX: 84
- EndNote: 79
Viewed (geographical distribution)
Country | # | Views | % |
---|
Total: | 0 |
HTML: | 0 |
PDF: | 0 |
XML: | 0 |
- 1